the optimized summed scored attributes method for the … · 2020. 3. 21. · 4 increases total...
TRANSCRIPT
Michael W. Kenyhercz, PhD Alexandra R. Klales, PhD Christopher W. Rainwater, MS Sara Fredette, BS
z
The Optimized Summed Scored Attributes Method for the
Classification of American Blacks and Whites: A Validation Study
Ancestry is difficult to estimate in forensic contexts
Pervasive use of non-metric traits in forensic
anthropology today
Hefner & Ousley (2014)
• Developed the OSSA method (Optimized Summed
Scored Attributes)
• 6 non-metric traits (Figure 1) ordinally scored each
trait score is converted to a binary score and then all
traits are summed
• Summed scores ≤ 3 are classified as Black and
summed scores ≥ 4 as White
• Method currently being used in forensic casework
Discussion & Conclusions
1 2 3 4 5
Table 1. Sample demographics by institution.
Sample
Methods
1 2 3 4 5
Introduction
Modern sample trait frequencies (DAFS, OCME) show more
overlap than the older HTH sample
Adjusting sectioning point to 4 increases total correct
classifications
Appears to reduce a White classification bias
Possible explanations:
Secular change
Increased admixture
Suggestions for application to modern forensic cases:
Adjust sectioning point to 4
Be familiar with the trait definitions and scores
Remain cognizant of range of human variation
Test the validity and reliability of the OSSA Method
Determine if a new sectioning point is more
appropriate for modern forensic cases
Research Goals
274 crania of American Black and White individuals
(Table 1)
• Historic: Hamann-Todd collection (n=208)
• Modern: positively identified forensic cases from the
Dept. Applied Forensic Sciences at Mercyhurst
University (n=28) and the New York City Office of the
Chief Medical Examiner (n=38)
Sexes pooled in accordance with Hefner & Ousley
(2014)
Six traits were scored based on the figures and
descriptions from Hefner (2009)
Trait frequencies tabulated
Classification accuracy tested using OSSA
Heuristic adjustment of cut-off scores to maximize
classification accuracy for the current sample
Results
Ancestry/Sex Group HTH DAFS OCME Total
Black Females (BF) 52 5 4 61
White Females (WF) 54 7 9 70
Black Males (BM) 50 0 6 56
White Males (WM) 52 16 19 87
Figure 1. Six traits from Hefner (2009) used in the OSSA Method developed by Hefner and Ousley (2014).
Results continued
SampleSectioning
Point
% Correct
Black
% Correct
White
Total Correct
Classification (%)
HTH 3 50.5 85.0 68.3
DAFS 3 20.0 82.6 71.4
OCME 3 70.0 96.4 89.5
HTH 4 80.2 69.2 77.9
DAFS 4 60.0 95.7 92.6
OCME 4 90.0 96.4 94.7
ANS INA IOB
1 2 3 1 2 3 4 5 1 2 3
B W B W B W B W B W B W B W B W B W B W B W
HTH 17.8 7.5 54.5 49.5 27.7 43.0 10.9 1.9 27.7 5.6 40.6 27.1 15.8 48.6 4.9 16.8 15.8 43.9 34.6 39.3 48.5 16.8
DAFS 0.0 0.0 100 15.0 0.0 85.0 25.0 0.0 25.0 0.0 25.0 10.0 25.0 50.0 0.0 40.0 25.0 30.0 25.0 65.0 50.0 5.0
OCME 54.5 12.9 27.3 35.5 18.2 51.6 18.2 0.0 18.2 0.0 45.5 3.2 18.2 41.9 0.0 54.8 9.1 41.9 36.4 54.8 54.5 3.2
NAW NBC PBD
1 2 3 0 1 2 3 4 0 1
B W B W B W B W B W B W B W B W B W B W
HTH 2.0 27.1 51.5 58.9 46.5 14.0 7.9 0.9 17.8 13.1 5.9 0.0 26.7 52.3 41.6 33.6 49.5 62.6 50.5 37.4
DAFS 0.0 50.0 75.0 50.0 25.0 0.0 0.0 0.0 0.0 0.0 25.0 20.0 50.0 55.0 25.0 25.0 75.0 90.0 25.0 10.0
OCME 18.2 58.1 54.5 41.9 27.3 0.0 72.7 0.0 9.1 12.9 0.0 3.2 9.1 61.3 9.1 22.6 81.8 96.8 18.2 3.2
Figure 2. Distribution of OSSA scores by
ancestry with each group’s mean shown
as a vertical checkered line.
Table 2. Trait frequencies by ancestry and sample.
OSSA score distributions
higher than reported by
Hefner & Ousley (2014)
(Figure 2)
Adjusting sectioning point to
4 increases total correct
classification in each sample
(Table 3)
Table 3. Correct classifications for each sample using the suggested (3) and adjusted (4) sectioning points.
zAcknowledgements
Thanks go to Dr. Dennis Dirkmaat of Mercyhurst University and Lyman Jellema of the Cleveland Museum of Natural
History for access to the collections used in this research. Thanks also go to Dr. Joseph Hefner for comments on this
research.
For a full list of references or a copy of the poster, contact: [email protected]